DOMAIN : Banking and Finance¶

1. Import and understand the data. [8 Marks]¶

In [1]:
 import numpy as np
 import pandas as pd
 from sklearn.tree import DecisionTreeClassifier

 import numpy as np
 import pandas as pd
 import seaborn as sb
 from matplotlib import pyplot as plt
 from sklearn.model_selection import train_test_split
 from sklearn.cluster import KMeans
 from sklearn import metrics

A. Import ‘BankChurners.csv’ as DataFrame. [1 Marks]

In [2]:
 bc=pd.read_csv('D:\\Great Learning\\FMT\\Project\\BC.csv')
 bc.head()
Out[2]:
CLIENTNUM Attrition_Flag Customer_Age Gender Dependent_count Education_Level Marital_Status Income_Category Card_Category Months_on_book ... Months_Inactive_12_mon Contacts_Count_12_mon Credit_Limit Total_Revolving_Bal Avg_Open_To_Buy Total_Amt_Chng_Q4_Q1 Total_Trans_Amt Total_Trans_Ct Total_Ct_Chng_Q4_Q1 Avg_Utilization_Ratio
0 768805383 Existing Customer 45 M 3 High School Married $60K - $80K Blue 39 ... 1 3 12691.0 777 11914.0 1.335 1144 42 1.625 0.061
1 818770008 Existing Customer 49 F 5 Graduate Single Less than $40K Blue 44 ... 1 2 8256.0 864 7392.0 1.541 1291 33 3.714 0.105
2 713982108 Existing Customer 51 M 3 Graduate Married $80K - $120K Blue 36 ... 1 0 3418.0 0 3418.0 2.594 1887 20 2.333 0.000
3 769911858 Existing Customer 40 F 4 High School NaN Less than $40K Blue 34 ... 4 1 3313.0 2517 796.0 1.405 1171 20 2.333 0.760
4 709106358 Existing Customer 40 M 3 Uneducated Married $60K - $80K Blue 21 ... 1 0 4716.0 0 4716.0 2.175 816 28 2.500 0.000

5 rows × 21 columns

B. Check the info of the dataset and also check if any duplicate records in the data. [2 Marks]

In [3]:
bc.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 10127 entries, 0 to 10126
Data columns (total 21 columns):
 #   Column                    Non-Null Count  Dtype  
---  ------                    --------------  -----  
 0   CLIENTNUM                 10127 non-null  int64  
 1   Attrition_Flag            10127 non-null  object 
 2   Customer_Age              10127 non-null  int64  
 3   Gender                    10127 non-null  object 
 4   Dependent_count           10127 non-null  int64  
 5   Education_Level           8608 non-null   object 
 6   Marital_Status            9378 non-null   object 
 7   Income_Category           10127 non-null  object 
 8   Card_Category             10127 non-null  object 
 9   Months_on_book            10127 non-null  int64  
 10  Total_Relationship_Count  10127 non-null  int64  
 11  Months_Inactive_12_mon    10127 non-null  int64  
 12  Contacts_Count_12_mon     10127 non-null  int64  
 13  Credit_Limit              10127 non-null  float64
 14  Total_Revolving_Bal       10127 non-null  int64  
 15  Avg_Open_To_Buy           10127 non-null  float64
 16  Total_Amt_Chng_Q4_Q1      10127 non-null  float64
 17  Total_Trans_Amt           10127 non-null  int64  
 18  Total_Trans_Ct            10127 non-null  int64  
 19  Total_Ct_Chng_Q4_Q1       10127 non-null  float64
 20  Avg_Utilization_Ratio     10127 non-null  float64
dtypes: float64(5), int64(10), object(6)
memory usage: 1.6+ MB
In [4]:
dup = bc[bc.duplicated()]
dup
Out[4]:
CLIENTNUM Attrition_Flag Customer_Age Gender Dependent_count Education_Level Marital_Status Income_Category Card_Category Months_on_book ... Months_Inactive_12_mon Contacts_Count_12_mon Credit_Limit Total_Revolving_Bal Avg_Open_To_Buy Total_Amt_Chng_Q4_Q1 Total_Trans_Amt Total_Trans_Ct Total_Ct_Chng_Q4_Q1 Avg_Utilization_Ratio

0 rows × 21 columns

In [5]:
bc.drop_duplicates()
Out[5]:
CLIENTNUM Attrition_Flag Customer_Age Gender Dependent_count Education_Level Marital_Status Income_Category Card_Category Months_on_book ... Months_Inactive_12_mon Contacts_Count_12_mon Credit_Limit Total_Revolving_Bal Avg_Open_To_Buy Total_Amt_Chng_Q4_Q1 Total_Trans_Amt Total_Trans_Ct Total_Ct_Chng_Q4_Q1 Avg_Utilization_Ratio
0 768805383 Existing Customer 45 M 3 High School Married $60K - $80K Blue 39 ... 1 3 12691.0 777 11914.0 1.335 1144 42 1.625 0.061
1 818770008 Existing Customer 49 F 5 Graduate Single Less than $40K Blue 44 ... 1 2 8256.0 864 7392.0 1.541 1291 33 3.714 0.105
2 713982108 Existing Customer 51 M 3 Graduate Married $80K - $120K Blue 36 ... 1 0 3418.0 0 3418.0 2.594 1887 20 2.333 0.000
3 769911858 Existing Customer 40 F 4 High School NaN Less than $40K Blue 34 ... 4 1 3313.0 2517 796.0 1.405 1171 20 2.333 0.760
4 709106358 Existing Customer 40 M 3 Uneducated Married $60K - $80K Blue 21 ... 1 0 4716.0 0 4716.0 2.175 816 28 2.500 0.000
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
10122 772366833 Existing Customer 50 M 2 Graduate Single $40K - $60K Blue 40 ... 2 3 4003.0 1851 2152.0 0.703 15476 117 0.857 0.462
10123 710638233 Attrited Customer 41 M 2 NaN Divorced $40K - $60K Blue 25 ... 2 3 4277.0 2186 2091.0 0.804 8764 69 0.683 0.511
10124 716506083 Attrited Customer 44 F 1 High School Married Less than $40K Blue 36 ... 3 4 5409.0 0 5409.0 0.819 10291 60 0.818 0.000
10125 717406983 Attrited Customer 30 M 2 Graduate NaN $40K - $60K Blue 36 ... 3 3 5281.0 0 5281.0 0.535 8395 62 0.722 0.000
10126 714337233 Attrited Customer 43 F 2 Graduate Married Less than $40K Silver 25 ... 2 4 10388.0 1961 8427.0 0.703 10294 61 0.649 0.189

10127 rows × 21 columns

There are no duplicate records in the dataframe

C. Print the 5-point summary of the data and share your insights on data distribution. [2 Marks]

In [6]:
bc.describe().T
Out[6]:
count mean std min 25% 50% 75% max
CLIENTNUM 10127.0 7.391776e+08 3.690378e+07 708082083.0 7.130368e+08 7.179264e+08 7.731435e+08 8.283431e+08
Customer_Age 10127.0 4.632596e+01 8.016814e+00 26.0 4.100000e+01 4.600000e+01 5.200000e+01 7.300000e+01
Dependent_count 10127.0 2.346203e+00 1.298908e+00 0.0 1.000000e+00 2.000000e+00 3.000000e+00 5.000000e+00
Months_on_book 10127.0 3.592841e+01 7.986416e+00 13.0 3.100000e+01 3.600000e+01 4.000000e+01 5.600000e+01
Total_Relationship_Count 10127.0 3.812580e+00 1.554408e+00 1.0 3.000000e+00 4.000000e+00 5.000000e+00 6.000000e+00
Months_Inactive_12_mon 10127.0 2.341167e+00 1.010622e+00 0.0 2.000000e+00 2.000000e+00 3.000000e+00 6.000000e+00
Contacts_Count_12_mon 10127.0 2.455317e+00 1.106225e+00 0.0 2.000000e+00 2.000000e+00 3.000000e+00 6.000000e+00
Credit_Limit 10127.0 8.631954e+03 9.088777e+03 1438.3 2.555000e+03 4.549000e+03 1.106750e+04 3.451600e+04
Total_Revolving_Bal 10127.0 1.162814e+03 8.149873e+02 0.0 3.590000e+02 1.276000e+03 1.784000e+03 2.517000e+03
Avg_Open_To_Buy 10127.0 7.469140e+03 9.090685e+03 3.0 1.324500e+03 3.474000e+03 9.859000e+03 3.451600e+04
Total_Amt_Chng_Q4_Q1 10127.0 7.599407e-01 2.192068e-01 0.0 6.310000e-01 7.360000e-01 8.590000e-01 3.397000e+00
Total_Trans_Amt 10127.0 4.404086e+03 3.397129e+03 510.0 2.155500e+03 3.899000e+03 4.741000e+03 1.848400e+04
Total_Trans_Ct 10127.0 6.485869e+01 2.347257e+01 10.0 4.500000e+01 6.700000e+01 8.100000e+01 1.390000e+02
Total_Ct_Chng_Q4_Q1 10127.0 7.122224e-01 2.380861e-01 0.0 5.820000e-01 7.020000e-01 8.180000e-01 3.714000e+00
Avg_Utilization_Ratio 10127.0 2.748936e-01 2.756915e-01 0.0 2.300000e-02 1.760000e-01 5.030000e-01 9.990000e-01

The customer age is between 26 to 73. All ages of people are eligible for credit card
The minimum credit limit is around 1400 rs and maximum is 34000 rs.
The average total transaction amount is approximately 4400rs
The for average utilization ratio, the median is closer to 75% (Q3) than 25% (Q1). So the data are left-skewed.

D. Print the unique values of all the categorical columns in the dataset. Share your observations. [2 Marks]

In [7]:
def preprocess():
 bc1= bc.select_dtypes(include=['object'])
 print(bc1.columns.unique())
 for col in bc1.columns:
     # to print categories name only
    print(bc1[col].value_counts())
In [8]:
preprocess()

Index(['Attrition_Flag', 'Gender', 'Education_Level', 'Marital_Status',
       'Income_Category', 'Card_Category'],
      dtype='object')
Existing Customer    8500
Attrited Customer    1627
Name: Attrition_Flag, dtype: int64
F    5358
M    4769
Name: Gender, dtype: int64
Graduate         3128
High School      2013
Uneducated       1487
College          1013
Post-Graduate     516
Doctorate         451
Name: Education_Level, dtype: int64
Married     4687
Single      3943
Divorced     748
Name: Marital_Status, dtype: int64
Less than $40K    3561
$40K - $60K       1790
$80K - $120K      1535
$60K - $80K       1402
abc               1112
$120K +            727
Name: Income_Category, dtype: int64
Blue        9436
Silver       555
Gold         116
Platinum      20
Name: Card_Category, dtype: int64

The female customers hold credit cards more compared to male customers
The Postgraduates and doctrates refrain from using credit cards compared to graduates who use it the most
Married people use credit cards more compared to single people
People with salary less than 40K use credit cards more compared to people who earn more than 40K
The blue card with lowest credit limit is issued to more customers based on their salary limit and platinum card is issued to elite customers

E.Check for the % of missing values in the dataset. Do not treat the missing values here. [1 Mark]

In [9]:
def miss():
 print(bc.isnull().sum())
In [10]:
miss()

CLIENTNUM                      0
Attrition_Flag                 0
Customer_Age                   0
Gender                         0
Dependent_count                0
Education_Level             1519
Marital_Status               749
Income_Category                0
Card_Category                  0
Months_on_book                 0
Total_Relationship_Count       0
Months_Inactive_12_mon         0
Contacts_Count_12_mon          0
Credit_Limit                   0
Total_Revolving_Bal            0
Avg_Open_To_Buy                0
Total_Amt_Chng_Q4_Q1           0
Total_Trans_Amt                0
Total_Trans_Ct                 0
Total_Ct_Chng_Q4_Q1            0
Avg_Utilization_Ratio          0
dtype: int64
In [11]:
def fillnan(): 
 bc["Education_Level"].fillna("Nan", inplace = True)
 bc["Marital_Status"].fillna("Nan", inplace = True)
 print(bc.Education_Level.value_counts())
 print(bc.Marital_Status.value_counts())
In [12]:
fillnan()

Graduate         3128
High School      2013
Nan              1519
Uneducated       1487
College          1013
Post-Graduate     516
Doctorate         451
Name: Education_Level, dtype: int64
Married     4687
Single      3943
Nan          749
Divorced     748
Name: Marital_Status, dtype: int64

2. Data Pre-processing. [ 3 Marks]¶

A. Drop redundant columns from the data. [1 Mark]

In [13]:
bc1 = bc.loc[:,~bc.apply(lambda x: x.duplicated(),axis=1).all()].copy()
bc1
Out[13]:
CLIENTNUM Attrition_Flag Customer_Age Gender Dependent_count Education_Level Marital_Status Income_Category Card_Category Months_on_book ... Months_Inactive_12_mon Contacts_Count_12_mon Credit_Limit Total_Revolving_Bal Avg_Open_To_Buy Total_Amt_Chng_Q4_Q1 Total_Trans_Amt Total_Trans_Ct Total_Ct_Chng_Q4_Q1 Avg_Utilization_Ratio
0 768805383 Existing Customer 45 M 3 High School Married $60K - $80K Blue 39 ... 1 3 12691.0 777 11914.0 1.335 1144 42 1.625 0.061
1 818770008 Existing Customer 49 F 5 Graduate Single Less than $40K Blue 44 ... 1 2 8256.0 864 7392.0 1.541 1291 33 3.714 0.105
2 713982108 Existing Customer 51 M 3 Graduate Married $80K - $120K Blue 36 ... 1 0 3418.0 0 3418.0 2.594 1887 20 2.333 0.000
3 769911858 Existing Customer 40 F 4 High School Nan Less than $40K Blue 34 ... 4 1 3313.0 2517 796.0 1.405 1171 20 2.333 0.760
4 709106358 Existing Customer 40 M 3 Uneducated Married $60K - $80K Blue 21 ... 1 0 4716.0 0 4716.0 2.175 816 28 2.500 0.000
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
10122 772366833 Existing Customer 50 M 2 Graduate Single $40K - $60K Blue 40 ... 2 3 4003.0 1851 2152.0 0.703 15476 117 0.857 0.462
10123 710638233 Attrited Customer 41 M 2 Nan Divorced $40K - $60K Blue 25 ... 2 3 4277.0 2186 2091.0 0.804 8764 69 0.683 0.511
10124 716506083 Attrited Customer 44 F 1 High School Married Less than $40K Blue 36 ... 3 4 5409.0 0 5409.0 0.819 10291 60 0.818 0.000
10125 717406983 Attrited Customer 30 M 2 Graduate Nan $40K - $60K Blue 36 ... 3 3 5281.0 0 5281.0 0.535 8395 62 0.722 0.000
10126 714337233 Attrited Customer 43 F 2 Graduate Married Less than $40K Silver 25 ... 2 4 10388.0 1961 8427.0 0.703 10294 61 0.649 0.189

10127 rows × 21 columns

No redundant columns in the dataframe

B. Encode the Existing and Attrited customers to 0 and 1 respectively in the ‘Attrition_Flag’ variable. [2 Marks]
Note: Please use inplace=True while replacing values to make permanent changes

In [14]:
def encodeatt():
 bc['Attrition_Flag'].replace({'Existing Customer':0, 'Attrited Customer':1}, inplace=True) 
 bc['Attrition_Flag'].info()
In [15]:
encodeatt()

<class 'pandas.core.series.Series'>
RangeIndex: 10127 entries, 0 to 10126
Series name: Attrition_Flag
Non-Null Count  Dtype
--------------  -----
10127 non-null  int64
dtypes: int64(1)
memory usage: 79.2 KB
In [16]:
bc.head()
Out[16]:
CLIENTNUM Attrition_Flag Customer_Age Gender Dependent_count Education_Level Marital_Status Income_Category Card_Category Months_on_book ... Months_Inactive_12_mon Contacts_Count_12_mon Credit_Limit Total_Revolving_Bal Avg_Open_To_Buy Total_Amt_Chng_Q4_Q1 Total_Trans_Amt Total_Trans_Ct Total_Ct_Chng_Q4_Q1 Avg_Utilization_Ratio
0 768805383 0 45 M 3 High School Married $60K - $80K Blue 39 ... 1 3 12691.0 777 11914.0 1.335 1144 42 1.625 0.061
1 818770008 0 49 F 5 Graduate Single Less than $40K Blue 44 ... 1 2 8256.0 864 7392.0 1.541 1291 33 3.714 0.105
2 713982108 0 51 M 3 Graduate Married $80K - $120K Blue 36 ... 1 0 3418.0 0 3418.0 2.594 1887 20 2.333 0.000
3 769911858 0 40 F 4 High School Nan Less than $40K Blue 34 ... 4 1 3313.0 2517 796.0 1.405 1171 20 2.333 0.760
4 709106358 0 40 M 3 Uneducated Married $60K - $80K Blue 21 ... 1 0 4716.0 0 4716.0 2.175 816 28 2.500 0.000

5 rows × 21 columns

3. Exploratory Data Analysis. [3 Marks]¶

A. Perform Univariate and Multivariate analysis on the data. Share your insights. [ 3 Marks]

Univariate Analysis

In [17]:
def univariate():
 plt.figure(1)
 bc.Attrition_Flag.value_counts(normalize=True).plot(kind = 'bar', title = "Attrition",figsize=(4, 3))
 plt.figure(2)
 bc.Income_Category.value_counts(normalize=True).plot(kind = 'bar', title = "Income_Category",color='red')
 plt.figure(3)
 bc.Education_Level.value_counts(normalize=True).plot(kind = 'bar', title = "Education_Level",color='green')
In [18]:
univariate()

Multivariate Analysis

In [19]:
def multivariate():
 cols = bc.describe().columns[1:]
 dtype_mapping = dict(bc.dtypes)
 numeric_cols = [ c for c in cols if dtype_mapping[c] !='O' ]
 print(numeric_cols)
 bc_num=bc.filter(['Attrition_Flag','Customer_Age', 'Dependent_count','Months_on_book', 'Total_Relationship_Count','Months_Inactive_12_mon','Contacts_Count_12_mon','Credit_Limit','Total_Revolving_Bal','Avg_Open_To_Buy','Total_Amt_Chng_Q4_Q1','Total_Trans_Amt','Total_Trans_Ct','Total_Ct_Chng_Q4_Q1', 'Avg_Utilization_Ratio'],axis=1)
 import seaborn as sb
 cols_to_plot=bc_num.columns[2:21].tolist() + ['Attrition_Flag'] 
 sb.pairplot(bc[cols_to_plot],hue="Attrition_Flag", palette='hls')
 plt.show()
 corr1=bc_num.corr()
 sb.heatmap(corr1, cmap=None, annot=True,annot_kws={"size":5})
In [20]:
multivariate()

['Attrition_Flag', 'Customer_Age', 'Dependent_count', 'Months_on_book', 'Total_Relationship_Count', 'Months_Inactive_12_mon', 'Contacts_Count_12_mon', 'Credit_Limit', 'Total_Revolving_Bal', 'Avg_Open_To_Buy', 'Total_Amt_Chng_Q4_Q1', 'Total_Trans_Amt', 'Total_Trans_Ct', 'Total_Ct_Chng_Q4_Q1', 'Avg_Utilization_Ratio']

Dependant count,Months_on_book,Total_Relationship_Count,Months_Inactive_12_mon,Contacts_Count_12_mon has least correlation with almost all coloumns so they can be dropped

Months_on_book and Customer Age are highly correlated.
Attrition is highly correlated with Months_Inactive_12_mon and Contacts_Count_12_mon
Total_Revolving_Bal and Avg_Utilization_Ratio are highly correlated

4. Data Preparation. [5 Marks]¶

A. Find the percentage of outliers, in each column of the data. [2 Marks]
Note: Do not treat outliers here as there will be outliers in real case scenario (in age, the total amount of transactions, number of transactions, etc) and we would want our model to learn the underlying pattern for such customers.

In [21]:
def outliers():
    plt.figure(1)
    sb.boxplot(x="Attrition_Flag",y="Avg_Utilization_Ratio",data=bc)
    plt.figure(2)
    sb.boxplot(x="Attrition_Flag",y="Total_Trans_Amt",data=bc)
    plt.figure(3)
    sb.boxplot(x="Attrition_Flag",y="Total_Trans_Ct",data=bc)
    plt.figure(4)
    sb.boxplot(x="Attrition_Flag",y="Customer_Age",data=bc)
    outliers1 = bc[bc['Customer_Age'] > bc['Customer_Age'].mean() + 3 * bc['Customer_Age'].std()]
    print(outliers1)
    outliers2 = bc[bc['Total_Trans_Ct'] > bc['Total_Trans_Ct'].mean() + 3 * bc['Total_Trans_Ct'].std()]
    print(outliers2)
    outliers3 = bc[bc['Total_Trans_Amt'] > bc['Total_Trans_Amt'].mean() + 3 * bc['Total_Trans_Amt'].std()]
    print(outliers3)
In [22]:
outliers()

     CLIENTNUM  Attrition_Flag  Customer_Age Gender  Dependent_count  \
251  715952883               0            73      M                0   

    Education_Level Marital_Status Income_Category Card_Category  \
251     High School        Married     $40K - $60K          Blue   

     Months_on_book  ...  Months_Inactive_12_mon  Contacts_Count_12_mon  \
251              36  ...                       3                      2   

     Credit_Limit  Total_Revolving_Bal  Avg_Open_To_Buy  Total_Amt_Chng_Q4_Q1  \
251        4469.0                 1125           3344.0                 1.363   

     Total_Trans_Amt  Total_Trans_Ct  Total_Ct_Chng_Q4_Q1  \
251             1765              34                1.615   

     Avg_Utilization_Ratio  
251                  0.252  

[1 rows x 21 columns]
      CLIENTNUM  Attrition_Flag  Customer_Age Gender  Dependent_count  \
9324  708163758               0            41      M                3   
9586  784868958               0            56      F                1   

     Education_Level Marital_Status Income_Category Card_Category  \
9324             Nan        Married         $120K +          Blue   
9586     High School        Married             abc          Blue   

      Months_on_book  ...  Months_Inactive_12_mon  Contacts_Count_12_mon  \
9324              33  ...                       4                      3   
9586              49  ...                       2                      1   

      Credit_Limit  Total_Revolving_Bal  Avg_Open_To_Buy  \
9324       34516.0                  638          33878.0   
9586       17542.0                 2517          15025.0   

      Total_Amt_Chng_Q4_Q1  Total_Trans_Amt  Total_Trans_Ct  \
9324                 0.724            13085             139   
9586                 0.800            13939             138   

      Total_Ct_Chng_Q4_Q1  Avg_Utilization_Ratio  
9324                0.675                  0.018  
9586                0.792                  0.143  

[2 rows x 21 columns]
       CLIENTNUM  Attrition_Flag  Customer_Age Gender  Dependent_count  \
9106   788709408               0            41      F                2   
9138   709108533               0            49      M                4   
9143   771062733               0            44      F                2   
9145   713552958               0            43      M                5   
9151   711814608               0            41      F                5   
...          ...             ...           ...    ...              ...   
10116  714109308               0            46      M                5   
10117  712503408               0            57      M                2   
10120  710841183               0            54      M                1   
10121  713899383               0            56      F                1   
10122  772366833               0            50      M                2   

      Education_Level Marital_Status Income_Category Card_Category  \
9106              Nan        Married             abc          Blue   
9138         Graduate        Married    $80K - $120K          Blue   
9143         Graduate         Single  Less than $40K          Blue   
9145         Graduate            Nan     $60K - $80K          Blue   
9151      High School        Married     $40K - $60K        Silver   
...               ...            ...             ...           ...   
10116         College         Single    $80K - $120K          Blue   
10117        Graduate        Married    $80K - $120K          Blue   
10120     High School         Single     $60K - $80K          Blue   
10121        Graduate         Single  Less than $40K          Blue   
10122        Graduate         Single     $40K - $60K          Blue   

       Months_on_book  ...  Months_Inactive_12_mon  Contacts_Count_12_mon  \
9106               21  ...                       3                      1   
9138               29  ...                       2                      3   
9143               31  ...                       2                      1   
9145               28  ...                       3                      1   
9151               36  ...                       1                      1   
...               ...  ...                     ...                    ...   
10116              36  ...                       2                      3   
10117              40  ...                       3                      4   
10120              34  ...                       2                      0   
10121              50  ...                       1                      4   
10122              40  ...                       2                      3   

       Credit_Limit  Total_Revolving_Bal  Avg_Open_To_Buy  \
9106        14746.0                    0          14746.0   
9138         9959.0                 2216           7743.0   
9143         4291.0                 1285           3006.0   
9145        12773.0                 1637          11136.0   
9151        19281.0                 1805          17476.0   
...             ...                  ...              ...   
10116       13187.0                 2241          10946.0   
10117       17925.0                 1909          16016.0   
10120       13940.0                 2109          11831.0   
10121        3688.0                  606           3082.0   
10122        4003.0                 1851           2152.0   

       Total_Amt_Chng_Q4_Q1  Total_Trans_Amt  Total_Trans_Ct  \
9106                  0.857            14771             127   
9138                  0.747            15139              90   
9143                  0.846            15511             101   
9145                  0.919            14833             114   
9151                  0.893            15423             110   
...                     ...              ...             ...   
10116                 0.689            15354             112   
10117                 0.712            17498             111   
10120                 0.660            15577             114   
10121                 0.570            14596             120   
10122                 0.703            15476             117   

       Total_Ct_Chng_Q4_Q1  Avg_Utilization_Ratio  
9106                 0.628                  0.000  
9138                 0.837                  0.223  
9143                 0.629                  0.299  
9145                 0.583                  0.128  
9151                 0.692                  0.094  
...                    ...                    ...  
10116                0.931                  0.170  
10117                0.820                  0.106  
10120                0.754                  0.151  
10121                0.791                  0.164  
10122                0.857                  0.462  

[391 rows x 21 columns]

B. Replace 'abc' values with 'np.nan' in Income_Category column. [1 Mark]

In [23]:
def replaceval():
 bc2=bc.replace('abc', 'Nan')
 print(bc2['Income_Category'].value_counts())
In [24]:
replaceval()

Less than $40K    3561
$40K - $60K       1790
$80K - $120K      1535
$60K - $80K       1402
Nan               1112
$120K +            727
Name: Income_Category, dtype: int64

C. Separate Target and Predictor variables. [2 Marks]

In [87]:
def targetpredict():
 global X,y,bc1
 bc1=bc.drop(['CLIENTNUM'],axis=1)
 X = bc1.iloc[:,1 :]
 y = bc1.iloc[:,0]
 return(X,y)
In [88]:
targetpredict()
Out[88]:
(       Customer_Age Gender  Dependent_count Education_Level Marital_Status  \
 0                45      M                3     High School        Married   
 1                49      F                5        Graduate         Single   
 2                51      M                3        Graduate        Married   
 3                40      F                4     High School            Nan   
 4                40      M                3      Uneducated        Married   
 ...             ...    ...              ...             ...            ...   
 10122            50      M                2        Graduate         Single   
 10123            41      M                2             Nan       Divorced   
 10124            44      F                1     High School        Married   
 10125            30      M                2        Graduate            Nan   
 10126            43      F                2        Graduate        Married   
 
       Income_Category Card_Category  Months_on_book  Total_Relationship_Count  \
 0         $60K - $80K          Blue              39                         5   
 1      Less than $40K          Blue              44                         6   
 2        $80K - $120K          Blue              36                         4   
 3      Less than $40K          Blue              34                         3   
 4         $60K - $80K          Blue              21                         5   
 ...               ...           ...             ...                       ...   
 10122     $40K - $60K          Blue              40                         3   
 10123     $40K - $60K          Blue              25                         4   
 10124  Less than $40K          Blue              36                         5   
 10125     $40K - $60K          Blue              36                         4   
 10126  Less than $40K        Silver              25                         6   
 
        Months_Inactive_12_mon  Contacts_Count_12_mon  Credit_Limit  \
 0                           1                      3       12691.0   
 1                           1                      2        8256.0   
 2                           1                      0        3418.0   
 3                           4                      1        3313.0   
 4                           1                      0        4716.0   
 ...                       ...                    ...           ...   
 10122                       2                      3        4003.0   
 10123                       2                      3        4277.0   
 10124                       3                      4        5409.0   
 10125                       3                      3        5281.0   
 10126                       2                      4       10388.0   
 
        Total_Revolving_Bal  Avg_Open_To_Buy  Total_Amt_Chng_Q4_Q1  \
 0                      777          11914.0                 1.335   
 1                      864           7392.0                 1.541   
 2                        0           3418.0                 2.594   
 3                     2517            796.0                 1.405   
 4                        0           4716.0                 2.175   
 ...                    ...              ...                   ...   
 10122                 1851           2152.0                 0.703   
 10123                 2186           2091.0                 0.804   
 10124                    0           5409.0                 0.819   
 10125                    0           5281.0                 0.535   
 10126                 1961           8427.0                 0.703   
 
        Total_Trans_Amt  Total_Trans_Ct  Total_Ct_Chng_Q4_Q1  \
 0                 1144              42                1.625   
 1                 1291              33                3.714   
 2                 1887              20                2.333   
 3                 1171              20                2.333   
 4                  816              28                2.500   
 ...                ...             ...                  ...   
 10122            15476             117                0.857   
 10123             8764              69                0.683   
 10124            10291              60                0.818   
 10125             8395              62                0.722   
 10126            10294              61                0.649   
 
        Avg_Utilization_Ratio  
 0                      0.061  
 1                      0.105  
 2                      0.000  
 3                      0.760  
 4                      0.000  
 ...                      ...  
 10122                  0.462  
 10123                  0.511  
 10124                  0.000  
 10125                  0.000  
 10126                  0.189  
 
 [10127 rows x 19 columns],
 0        0
 1        0
 2        0
 3        0
 4        0
         ..
 10122    0
 10123    1
 10124    1
 10125    1
 10126    1
 Name: Attrition_Flag, Length: 10127, dtype: int64)

5. Split data into train, validation and test set. [2 Marks]¶

A. Split data into 2 parts, say temporary and test. Please use below line of code. [1 Mark]
X_temp, X_test, y_temp, y_test = train_test_split(X, y, test_size=0.2, random_state=1, stratify=y)

In [27]:
X_temp, X_test, y_temp, y_test = train_test_split(X, y, test_size=0.2, random_state=1, stratify=y)

B. Now split the temporary set into train and validation. Please use below line of code. [1 Mark]
X_train, X_val, y_train, y_val = train_test_split(X_temp, y_temp, test_size=0.25, random_state=1, stratify=y_temp)

In [28]:
X_train, X_val, y_train, y_val = train_test_split(X_temp, y_temp, test_size=0.25, random_state=1, stratify=y_temp)
In [29]:
X_train.Income_Category.value_counts()
Out[29]:
Less than $40K    2129
$40K - $60K       1059
$80K - $120K       953
$60K - $80K        831
abc                654
$120K +            449
Name: Income_Category, dtype: int64
In [30]:
X_test.Income_Category.value_counts()
Out[30]:
Less than $40K    696
$40K - $60K       370
$60K - $80K       292
$80K - $120K      289
abc               237
$120K +           142
Name: Income_Category, dtype: int64
In [31]:
X_val.Income_Category.value_counts()
Out[31]:
Less than $40K    736
$40K - $60K       361
$80K - $120K      293
$60K - $80K       279
abc               221
$120K +           136
Name: Income_Category, dtype: int64
In [32]:
X_train.Education_Level.value_counts()
Out[32]:
Graduate         1854
High School      1228
Nan               928
Uneducated        881
College           618
Post-Graduate     312
Doctorate         254
Name: Education_Level, dtype: int64
In [33]:
X_test.Education_Level.value_counts()
Out[33]:
Graduate         651
High School      381
Uneducated       300
Nan              297
College          196
Post-Graduate    103
Doctorate         98
Name: Education_Level, dtype: int64
In [34]:
X_val.Education_Level.value_counts()
Out[34]:
Graduate         623
High School      404
Uneducated       306
Nan              294
College          199
Post-Graduate    101
Doctorate         99
Name: Education_Level, dtype: int64
In [35]:
X_train.Marital_Status.value_counts()
Out[35]:
Married     2819
Single      2369
Nan          457
Divorced     430
Name: Marital_Status, dtype: int64
In [36]:
X_test.Marital_Status.value_counts()
Out[36]:
Married     908
Single      804
Divorced    162
Nan         152
Name: Marital_Status, dtype: int64
In [37]:
X_val.Marital_Status.value_counts()
Out[37]:
Married     960
Single      770
Divorced    156
Nan         140
Name: Marital_Status, dtype: int64
In [38]:
y_train.value_counts()
Out[38]:
0    5099
1     976
Name: Attrition_Flag, dtype: int64
In [39]:
def traintest():
  X_temp, X_test, y_temp, y_test = train_test_split(X, y, test_size=0.2, random_state=1, stratify=y) 
  X_train, X_val, y_train, y_val = train_test_split(X_temp, y_temp, test_size=0.25, random_state=1, stratify=y_temp)

6. Missing value Imputation. [3 Marks]¶

A. Impute all missing values in "Education_Level", "Marital_Status", "Income_Category" columns using suitable techniques from X_train, X_val, X_test separately. [3 Marks]
Note: "Education_Level", "Marital_Status", "Income_Category" are categorical variables.

In [40]:
def Impute():
 from sklearn.impute import SimpleImputer
 ic_imputer = SimpleImputer(missing_values="Nan", strategy="most_frequent")
 X_train.Income_Category = ic_imputer.fit_transform(X_train["Income_Category"].values.reshape(-1,1))[:,0]
 X_test.Income_Category = ic_imputer.fit_transform(X_test["Income_Category"].values.reshape(-1,1))[:,0]
 X_val.Income_Category = ic_imputer.fit_transform(X_val["Income_Category"].values.reshape(-1,1))[:,0]
 X_train.Education_Level = ic_imputer.fit_transform(X_train["Education_Level"].values.reshape(-1,1))[:,0]
 X_test.Education_Level = ic_imputer.fit_transform(X_test["Education_Level"].values.reshape(-1,1))[:,0]
 X_val.Education_Level = ic_imputer.fit_transform(X_val["Education_Level"].values.reshape(-1,1))[:,0]
 X_train.Marital_Status = ic_imputer.fit_transform(X_train["Marital_Status"].values.reshape(-1,1))[:,0]
 X_test.Marital_Status = ic_imputer.fit_transform(X_test["Marital_Status"].values.reshape(-1,1))[:,0]
 X_val.Marital_Status = ic_imputer.fit_transform(X_val["Marital_Status"].values.reshape(-1,1))[:,0]
 return()
In [42]:
from sklearn.impute import SimpleImputer
ic_imputer = SimpleImputer(missing_values="Nan", strategy="most_frequent")
X_test.Income_Category = ic_imputer.fit_transform(X_test["Income_Category"].values.reshape(-1,1))[:,0]
X_test.Income_Category.value_counts()
Out[42]:
Less than $40K    696
$40K - $60K       370
$60K - $80K       292
$80K - $120K      289
abc               237
$120K +           142
Name: Income_Category, dtype: int64
In [43]:
from sklearn.impute import SimpleImputer
ic_imputer = SimpleImputer(missing_values="Nan", strategy="most_frequent")
X_val.Income_Category = ic_imputer.fit_transform(X_val["Income_Category"].values.reshape(-1,1))[:,0]
X_val.Income_Category.value_counts()
Out[43]:
Less than $40K    736
$40K - $60K       361
$80K - $120K      293
$60K - $80K       279
abc               221
$120K +           136
Name: Income_Category, dtype: int64
In [44]:
from sklearn.impute import SimpleImputer
ic_imputer = SimpleImputer(missing_values="Nan", strategy="most_frequent")
X_train.Education_Level = ic_imputer.fit_transform(X_train["Education_Level"].values.reshape(-1,1))[:,0]
X_train.Education_Level.value_counts()
Out[44]:
Graduate         2782
High School      1228
Uneducated        881
College           618
Post-Graduate     312
Doctorate         254
Name: Education_Level, dtype: int64
In [45]:
from sklearn.impute import SimpleImputer
ic_imputer = SimpleImputer(missing_values="Nan", strategy="most_frequent")
X_test.Education_Level = ic_imputer.fit_transform(X_test["Education_Level"].values.reshape(-1,1))[:,0]
X_test.Education_Level.value_counts()
Out[45]:
Graduate         948
High School      381
Uneducated       300
College          196
Post-Graduate    103
Doctorate         98
Name: Education_Level, dtype: int64
In [46]:
from sklearn.impute import SimpleImputer
ic_imputer = SimpleImputer(missing_values="Nan", strategy="most_frequent")
X_val.Education_Level = ic_imputer.fit_transform(X_val["Education_Level"].values.reshape(-1,1))[:,0]
X_val.Education_Level.value_counts()
Out[46]:
Graduate         917
High School      404
Uneducated       306
College          199
Post-Graduate    101
Doctorate         99
Name: Education_Level, dtype: int64
In [47]:
from sklearn.impute import SimpleImputer
ic_imputer = SimpleImputer(missing_values="Nan", strategy="most_frequent")
X_train.Marital_Status = ic_imputer.fit_transform(X_train["Marital_Status"].values.reshape(-1,1))[:,0]
X_train.Marital_Status.value_counts()
Out[47]:
Married     3276
Single      2369
Divorced     430
Name: Marital_Status, dtype: int64
In [48]:
from sklearn.impute import SimpleImputer
ic_imputer = SimpleImputer(missing_values="Nan", strategy="most_frequent")
X_test.Marital_Status = ic_imputer.fit_transform(X_test["Marital_Status"].values.reshape(-1,1))[:,0]
X_test.Marital_Status.value_counts()
Out[48]:
Married     1060
Single       804
Divorced     162
Name: Marital_Status, dtype: int64
In [49]:
from sklearn.impute import SimpleImputer
ic_imputer = SimpleImputer(missing_values="Nan", strategy="most_frequent")
X_val.Marital_Status = ic_imputer.fit_transform(X_val["Marital_Status"].values.reshape(-1,1))[:,0]
X_val.Marital_Status.value_counts()
Out[49]:
Married     1100
Single       770
Divorced     156
Name: Marital_Status, dtype: int64

7. Encoding Categorical columns. [3 Marks]¶

A. Encode all Categorical columns in X_train, X_val, X_test using one-hot encoding. [ 3 Marks]

In [50]:
X_train.shape
Out[50]:
(6075, 19)
In [51]:
X_test.shape
Out[51]:
(2026, 19)
In [52]:
X_val.shape
Out[52]:
(2026, 19)
In [53]:
y_train.shape
Out[53]:
(6075,)

7. Encoding Categorical columns. [3 Marks]¶

A. Encode all Categorical columns in X_train, X_val, X_test using one-hot encoding. [ 3 Marks]

In [54]:
from sklearn import preprocessing
le = preprocessing.LabelEncoder()
#entel = pd.DataFrame()
object_col = X_train.columns[pd.Series(X_train.columns).apply(lambda x : X_train[x].dtype == 'O')] 
for i in X_train.columns:
    if i in object_col:
        X_train[i] = le.fit_transform(X_train[i])
    else:
        X_train[i] = X_train[i]
In [55]:
X_train1=X_train.copy()
X_train1
Out[55]:
Customer_Age Gender Dependent_count Education_Level Marital_Status Income_Category Card_Category Months_on_book Total_Relationship_Count Months_Inactive_12_mon Contacts_Count_12_mon Credit_Limit Total_Revolving_Bal Avg_Open_To_Buy Total_Amt_Chng_Q4_Q1 Total_Trans_Amt Total_Trans_Ct Total_Ct_Chng_Q4_Q1 Avg_Utilization_Ratio
800 40 1 2 2 2 0 0 21 6 4 3 20056.0 1602 18454.0 0.466 1687 46 0.533 0.080
498 44 1 1 2 1 5 0 34 6 2 0 2885.0 1895 990.0 0.387 1366 31 0.632 0.657
4356 48 1 4 3 1 3 0 36 5 1 2 6798.0 2517 4281.0 0.873 4327 79 0.881 0.370
407 41 1 2 2 1 2 3 36 6 2 0 27000.0 0 27000.0 0.610 1209 39 0.300 0.000
8728 46 1 4 3 0 1 3 36 2 2 3 15034.0 1356 13678.0 0.754 7737 84 0.750 0.090
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
987 47 0 2 2 2 4 0 36 6 2 2 5145.0 0 5145.0 0.693 1588 35 0.944 0.000
5633 47 1 3 5 2 0 0 36 3 3 4 34516.0 0 34516.0 0.517 2020 33 0.375 0.000
9621 50 0 0 4 0 4 0 36 3 4 1 3468.0 2221 1247.0 0.783 14222 122 0.627 0.640
593 45 0 4 5 1 4 0 38 3 4 3 4908.0 0 4908.0 0.993 827 21 0.750 0.000
4035 55 0 2 0 1 5 0 41 6 2 2 7613.0 0 7613.0 0.680 4268 69 0.725 0.000

6075 rows × 19 columns

In [56]:
object_col = X_test.columns[pd.Series(X_test.columns).apply(lambda x : X_test[x].dtype == 'O')] 
for i in X_test.columns:
    if i in object_col:
        X_test[i] = le.fit_transform(X_test[i])
    else:
        X_test[i] = X_test[i]
In [57]:
X_test
Out[57]:
Customer_Age Gender Dependent_count Education_Level Marital_Status Income_Category Card_Category Months_on_book Total_Relationship_Count Months_Inactive_12_mon Contacts_Count_12_mon Credit_Limit Total_Revolving_Bal Avg_Open_To_Buy Total_Amt_Chng_Q4_Q1 Total_Trans_Amt Total_Trans_Ct Total_Ct_Chng_Q4_Q1 Avg_Utilization_Ratio
9760 32 1 1 3 2 3 0 26 2 3 2 6407.0 1130 5277.0 0.756 14471 93 0.603 0.176
7413 50 1 1 4 2 2 0 36 4 3 2 2317.0 0 2317.0 0.734 2214 41 0.519 0.000
6074 54 0 2 3 1 1 0 36 3 3 3 3892.0 0 3892.0 0.738 4318 74 0.762 0.000
3520 61 1 0 5 1 0 0 36 4 3 4 24172.0 2517 21655.0 0.424 1658 27 0.500 0.104
6103 41 0 3 0 1 1 0 15 5 3 4 4312.0 2517 1795.0 0.741 2693 56 0.436 0.584
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
1484 32 1 0 4 1 2 0 26 6 2 3 5478.0 1491 3987.0 1.304 2456 55 0.571 0.272
1271 32 0 0 3 0 4 0 36 5 2 4 2699.0 1609 1090.0 0.893 2141 56 0.806 0.596
4002 45 1 2 2 1 0 0 40 4 4 4 16476.0 1374 15102.0 0.459 4717 71 0.690 0.083
8562 33 0 4 0 1 1 0 26 1 2 3 2874.0 2517 357.0 0.730 4794 88 0.600 0.876
413 55 0 2 2 1 4 0 37 3 3 1 2453.0 968 1485.0 1.081 2218 50 0.613 0.395

2026 rows × 19 columns

In [58]:
object_col = X_val.columns[pd.Series(X_val.columns).apply(lambda x : X_val[x].dtype == 'O')] 
for i in X_val.columns:
    if i in object_col:
        X_val[i] = le.fit_transform(X_val[i])
    else:
        X_val[i] = X_val[i]
In [59]:
X_val
Out[59]:
Customer_Age Gender Dependent_count Education_Level Marital_Status Income_Category Card_Category Months_on_book Total_Relationship_Count Months_Inactive_12_mon Contacts_Count_12_mon Credit_Limit Total_Revolving_Bal Avg_Open_To_Buy Total_Amt_Chng_Q4_Q1 Total_Trans_Amt Total_Trans_Ct Total_Ct_Chng_Q4_Q1 Avg_Utilization_Ratio
2894 37 1 0 4 2 3 0 27 5 2 3 15326.0 0 15326.0 1.159 2990 55 0.964 0.000
9158 58 1 2 5 2 3 0 46 1 3 1 10286.0 0 10286.0 0.908 8199 59 0.903 0.000
9618 42 1 3 5 1 0 2 23 3 4 3 34516.0 2070 32446.0 0.880 13781 102 0.545 0.060
9910 47 1 3 2 1 3 0 36 3 2 3 9683.0 1116 8567.0 0.721 15627 104 0.825 0.115
5497 60 0 1 3 2 4 0 36 5 2 2 2688.0 1617 1071.0 0.552 4183 71 0.614 0.602
... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ... ...
426 39 1 2 5 2 2 0 36 5 3 3 13961.0 1299 12662.0 0.842 783 20 0.667 0.093
5091 53 0 0 2 2 4 0 40 4 1 4 1438.3 0 1438.3 0.503 2014 53 0.472 0.000
500 45 0 4 2 1 5 0 35 6 3 0 2801.0 2043 758.0 0.583 1241 46 0.643 0.729
353 53 0 2 2 2 1 0 42 6 2 0 5332.0 0 5332.0 0.939 1371 39 1.167 0.000
6319 38 0 0 2 2 4 0 32 6 3 3 1755.0 750 1005.0 0.659 5000 79 0.837 0.427

2026 rows × 19 columns

In [60]:
y_train
Out[60]:
800     0
498     0
4356    0
407     0
8728    0
       ..
987     0
5633    1
9621    0
593     1
4035    0
Name: Attrition_Flag, Length: 6075, dtype: int64

8. Model Building [22 Marks]¶

A. Build a Decision Tree, Logistic regression, Random Forest, Gradient Boost and XG Boost model on train data. [ 5 Marks]

In [61]:
from sklearn.tree import DecisionTreeRegressor
from sklearn.metrics import confusion_matrix
dt = DecisionTreeRegressor(random_state=1)
dt.fit(X_train, y_train)
y_pred = dt.predict(X_test)

tn, fp, fn, tp = confusion_matrix(y_test, y_pred).ravel()

dt_score = dt.score(X_test, y_test)
dt_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
dt_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
dt_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['DecisionTree'], 'Accuracy' : [dt_score], 'Precision' : [dt_precision],
                      'True positive rate' : [dt_recall], 'True negative rate' : [dt_specificity],
                       'False positive rate' :  [1-dt_specificity]})

result
Out[61]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 DecisionTree 0.475932 0.765 0.809 0.952 0.048
In [62]:
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import recall_score
lr = LogisticRegression(solver='liblinear', random_state=0)

lr.fit(X_train, y_train)

pred = lr.predict(X_test)
tn, fp, fn, tp = confusion_matrix(y_test,pred).ravel()

lr_score = lr.score(X_test, y_test)
lr_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
lr_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
lr_specificity =round(tn/(tn+fp),3)    # Of all the people without Chds, how many were recognised to not have Chd

result = pd.DataFrame({'Model' : ['Logistic Regression'], 'Accuracy' : [lr_score], 'Precision' : [lr_precision],
                      'True positive rate' : [lr_recall], 'True negative rate' : [lr_specificity],
                       'False positive rate' :  [1-lr_specificity]})
result
Out[62]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 Logistic Regression 0.890424 0.744 0.483 0.968 0.032
In [63]:
from sklearn.ensemble import RandomForestClassifier

rf = RandomForestClassifier(random_state=1)

rf.fit(X_train, y_train)

pred = rf.predict(X_test)

tn, fp, fn, tp = confusion_matrix(y_test, pred).ravel()

rf_score = rf.score(X_test, y_test)
rf_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
rf_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
rf_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['RandomForest'], 'Accuracy' : [rf_score], 'Precision' : [rf_precision],
                      'True positive rate' : [rf_recall], 'True negative rate' : [rf_specificity],
                       'False positive rate' :  [1-rf_specificity]})
result
Out[63]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 RandomForest 0.960513 0.907 0.84 0.984 0.016
In [64]:
from sklearn.ensemble import GradientBoostingClassifier

 
gb = GradientBoostingClassifier(n_estimators=500,learning_rate=0.05,random_state=100,max_features=5)
gb.fit(X_train, y_train)
pred = gb.predict(X_test)

tn, fp, fn, tp = confusion_matrix(y_test, pred).ravel()

gb_score = gb.score(X_test, y_test)
gb_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
gb_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
gb_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['GradientBoosting'], 'Accuracy' : [gb_score], 'Precision' : [gb_precision],
                      'True positive rate' : [gb_recall], 'True negative rate' : [gb_specificity],
                       'False positive rate' :  [1-gb_specificity]})
result
#print("mse Score(gb) : ",mse(y_test, y_pred))
Out[64]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 GradientBoosting 0.977789 0.943 0.917 0.989 0.011
In [65]:
from xgboost import XGBClassifier
xg = XGBClassifier(learning_rate=0.001,max_depth = 1,n_estimators = 100,scale_pos_weight=5)
xg.fit(X_train, y_train)

pred = xg.predict(X_test)

tn, fp, fn, tp = confusion_matrix(y_test, pred).ravel()

xg_score = xg.score(X_test, y_test)
xg_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
xg_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
xg_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['XGBClassifier'], 'Accuracy' : [xg_score], 'Precision' : [xg_precision],
                      'True positive rate' : [xg_recall], 'True negative rate' : [xg_specificity],
                       'False positive rate' :  [1-xg_specificity]})
result
Out[65]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 XGBClassifier 0.733465 0.355 0.806 0.72 0.28

B. Validate the models built above on Validation set. [ 4 Marks]

In [66]:
y_pred = dt.predict(X_val)

tn, fp, fn, tp = confusion_matrix(y_val, y_pred).ravel()

dt_score = dt.score(X_val, y_val)
dt_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
dt_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
dt_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['DecisionTree'], 'Accuracy' : [dt_score], 'Precision' : [dt_precision],
                      'True positive rate' : [dt_recall], 'True negative rate' : [dt_specificity],
                       'False positive rate' :  [1-dt_specificity]})

result
Out[66]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 DecisionTree 0.546691 0.804 0.819 0.962 0.038
In [67]:
pred = lr.predict(X_val)
tn, fp, fn, tp = confusion_matrix(y_val,pred).ravel()

lr_score = lr.score(X_val, y_val)
lr_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
lr_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
lr_specificity =round(tn/(tn+fp),3)    # Of all the people without Chds, how many were recognised to not have Chd

result = pd.DataFrame({'Model' : ['Logistic Regression'], 'Accuracy' : [lr_score], 'Precision' : [lr_precision],
                      'True positive rate' : [lr_recall], 'True negative rate' : [lr_specificity],
                       'False positive rate' :  [1-lr_specificity]})
result
Out[67]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 Logistic Regression 0.897828 0.764 0.528 0.969 0.031
In [68]:
pred = rf.predict(X_val)

tn, fp, fn, tp = confusion_matrix(y_val, pred).ravel()

rf_score = rf.score(X_val, y_val)
rf_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
rf_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
rf_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['RandomForest'], 'Accuracy' : [rf_score], 'Precision' : [rf_precision],
                      'True positive rate' : [rf_recall], 'True negative rate' : [rf_specificity],
                       'False positive rate' :  [1-rf_specificity]})
result
Out[68]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 RandomForest 0.962488 0.931 0.828 0.988 0.012
In [69]:
pred = gb.predict(X_val)

tn, fp, fn, tp = confusion_matrix(y_val, pred).ravel()

gb_score = gb.score(X_val, y_val)
gb_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
gb_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
gb_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['GradientBoosting'], 'Accuracy' : [gb_score], 'Precision' : [gb_precision],
                      'True positive rate' : [gb_recall], 'True negative rate' : [gb_specificity],
                       'False positive rate' :  [1-gb_specificity]})
result
Out[69]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 GradientBoosting 0.97384 0.945 0.89 0.99 0.01
In [70]:
pred = xg.predict(X_val)

tn, fp, fn, tp = confusion_matrix(y_val, pred).ravel()

xg_score = xg.score(X_val, y_val)
xg_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
xg_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
xg_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['XGBClassifier'], 'Accuracy' : [xg_score], 'Precision' : [xg_precision],
                      'True positive rate' : [xg_recall], 'True negative rate' : [xg_specificity],
                       'False positive rate' :  [1-xg_specificity]})
result
Out[70]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 XGBClassifier 0.74383 0.369 0.837 0.726 0.274

C. Mention which model is giving us better recall score. [1 Mark]

The Gradient Boosting model gives highest recall compared to all other models

D. Try target balancing with Up sampling the data. Build all the above models on this balanced data. Share your insights on the recall score of the models. [ 4 Marks]

In [71]:
y_train.value_counts()
Out[71]:
0    5099
1     976
Name: Attrition_Flag, dtype: int64
In [72]:
train = pd.concat([X_train, y_train], axis = 1)

extra_samples = train[train['Attrition_Flag'] == 1].sample(n = 900, random_state = 1)

train_upsampled = pd.concat([train, extra_samples], axis = 0)
In [73]:
X_train = train_upsampled.drop(columns = 'Attrition_Flag')
y_train = train_upsampled.Attrition_Flag

X_train.shape

y_train.value_counts()
Out[73]:
0    5099
1    1876
Name: Attrition_Flag, dtype: int64
In [74]:
dt.fit(X_train, y_train)
y_pred = dt.predict(X_test)

tn, fp, fn, tp = confusion_matrix(y_test, y_pred).ravel()

dt_score = dt.score(X_test, y_test)
dt_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
dt_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
dt_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['DecisionTree'], 'Accuracy' : [dt_score], 'Precision' : [dt_precision],
                      'True positive rate' : [dt_recall], 'True negative rate' : [dt_specificity],
                       'False positive rate' :  [1-dt_specificity]})

result
Out[74]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 DecisionTree 0.494256 0.771 0.818 0.954 0.046
In [75]:
lr.fit(X_train, y_train)

pred = lr.predict(X_test)
tn, fp, fn, tp = confusion_matrix(y_test,pred).ravel()

lr_score = lr.score(X_test, y_test)
lr_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
lr_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
lr_specificity =round(tn/(tn+fp),3)    # Of all the people without Chds, how many were recognised to not have Chd

result = pd.DataFrame({'Model' : ['Logistic Regression'], 'Accuracy' : [lr_score], 'Precision' : [lr_precision],
                      'True positive rate' : [lr_recall], 'True negative rate' : [lr_specificity],
                       'False positive rate' :  [1-lr_specificity]})
result
Out[75]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 Logistic Regression 0.882034 0.626 0.658 0.925 0.075
In [76]:
rf.fit(X_train, y_train)
pred = rf.predict(X_test)

tn, fp, fn, tp = confusion_matrix(y_test, pred).ravel()

rf_score = rf.score(X_test, y_test)
rf_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
rf_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
rf_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['RandomForest'], 'Accuracy' : [rf_score], 'Precision' : [rf_precision],
                      'True positive rate' : [rf_recall], 'True negative rate' : [rf_specificity],
                       'False positive rate' :  [1-rf_specificity]})
result
Out[76]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 RandomForest 0.962981 0.896 0.871 0.981 0.019
In [77]:
gb.fit(X_train, y_train)
pred = gb.predict(X_test)

tn, fp, fn, tp = confusion_matrix(y_test, pred).ravel()

gb_score = gb.score(X_test, y_test)
gb_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
gb_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
gb_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['GradientBoosting'], 'Accuracy' : [gb_score], 'Precision' : [gb_precision],
                      'True positive rate' : [gb_recall], 'True negative rate' : [gb_specificity],
                       'False positive rate' :  [1-gb_specificity]})
result
Out[77]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 GradientBoosting 0.977295 0.921 0.938 0.985 0.015
In [78]:
xg.fit(X_train, y_train)

pred = xg.predict(X_test)

tn, fp, fn, tp = confusion_matrix(y_test, pred).ravel()

xg_score = xg.score(X_test, y_test)
xg_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
xg_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
xg_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['XGBClassifier'], 'Accuracy' : [xg_score], 'Precision' : [xg_precision],
                      'True positive rate' : [xg_recall], 'True negative rate' : [xg_specificity],
                       'False positive rate' :  [1-xg_specificity]})
result
Out[78]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 XGBClassifier 0.7231 0.345 0.809 0.707 0.293

E. Try target balancing with Down sampling the data. Build all the above models on this balanced data. Share your insights on the recall score of the models. [ 4 Marks]

In [84]:
bc1.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 10127 entries, 0 to 10126
Data columns (total 21 columns):
 #   Column                    Non-Null Count  Dtype  
---  ------                    --------------  -----  
 0   CLIENTNUM                 10127 non-null  int64  
 1   Attrition_Flag            10127 non-null  object 
 2   Customer_Age              10127 non-null  int64  
 3   Gender                    10127 non-null  object 
 4   Dependent_count           10127 non-null  int64  
 5   Education_Level           10127 non-null  object 
 6   Marital_Status            10127 non-null  object 
 7   Income_Category           10127 non-null  object 
 8   Card_Category             10127 non-null  object 
 9   Months_on_book            10127 non-null  int64  
 10  Total_Relationship_Count  10127 non-null  int64  
 11  Months_Inactive_12_mon    10127 non-null  int64  
 12  Contacts_Count_12_mon     10127 non-null  int64  
 13  Credit_Limit              10127 non-null  float64
 14  Total_Revolving_Bal       10127 non-null  int64  
 15  Avg_Open_To_Buy           10127 non-null  float64
 16  Total_Amt_Chng_Q4_Q1      10127 non-null  float64
 17  Total_Trans_Amt           10127 non-null  int64  
 18  Total_Trans_Ct            10127 non-null  int64  
 19  Total_Ct_Chng_Q4_Q1       10127 non-null  float64
 20  Avg_Utilization_Ratio     10127 non-null  float64
dtypes: float64(5), int64(10), object(6)
memory usage: 1.6+ MB
In [89]:
d1ind = bc1[bc1['Attrition_Flag'] == 0].index   
d1 = len(bc1[bc1['Attrition_Flag'] == 0])             
print(d1)

d2ind = bc1[bc1['Attrition_Flag'] == 1].index   
d2 = len(bc1[bc1['Attrition_Flag'] == 1])             
print(d2)

8500
1627
In [90]:
random_indices = np.random.choice(d1ind, d1 - 6500 , replace=False)    #Randomly pick up 200 non-diab indices
In [91]:
down_sample_indices = np.concatenate([d2ind,random_indices])
In [92]:
bc1_down_sample = bc1.loc[down_sample_indices]  # Extract all those records for diab and non-diab to create new set
bc1_down_sample.shape
bc1_down_sample.groupby(["Attrition_Flag"]).count()
Out[92]:
Customer_Age Gender Dependent_count Education_Level Marital_Status Income_Category Card_Category Months_on_book Total_Relationship_Count Months_Inactive_12_mon Contacts_Count_12_mon Credit_Limit Total_Revolving_Bal Avg_Open_To_Buy Total_Amt_Chng_Q4_Q1 Total_Trans_Amt Total_Trans_Ct Total_Ct_Chng_Q4_Q1 Avg_Utilization_Ratio
Attrition_Flag
0 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000 2000
1 1627 1627 1627 1627 1627 1627 1627 1627 1627 1627 1627 1627 1627 1627 1627 1627 1627 1627 1627
In [93]:
X1 = bc1_down_sample.iloc[:,1 :]
y1= bc1_down_sample.iloc[:,0]
In [94]:
from sklearn.impute import SimpleImputer
ic_imputer = SimpleImputer(missing_values="Nan", strategy="most_frequent")
X1.Income_Category = ic_imputer.fit_transform(X1["Income_Category"].values.reshape(-1,1))[:,0]
X1.Income_Category.value_counts()
Out[94]:
Less than $40K    1273
$40K - $60K        644
$80K - $120K       539
$60K - $80K        468
abc                414
$120K +            289
Name: Income_Category, dtype: int64
In [95]:
from sklearn.impute import SimpleImputer
ic_imputer = SimpleImputer(missing_values="Nan", strategy="most_frequent")
X1.Education_Level = ic_imputer.fit_transform(X1["Education_Level"].values.reshape(-1,1))[:,0]
X1.Education_Level.value_counts()
Out[95]:
Graduate         1644
High School       715
Uneducated        533
College           345
Post-Graduate     206
Doctorate         184
Name: Education_Level, dtype: int64
In [96]:
from sklearn.impute import SimpleImputer
ic_imputer = SimpleImputer(missing_values="Nan", strategy="most_frequent")
X1.Marital_Status = ic_imputer.fit_transform(X1["Marital_Status"].values.reshape(-1,1))[:,0]
X1.Marital_Status.value_counts()
Out[96]:
Married     1908
Single      1434
Divorced     285
Name: Marital_Status, dtype: int64
In [97]:
from sklearn import preprocessing
le = preprocessing.LabelEncoder()
#entel = pd.DataFrame()
object_col = X1.columns[pd.Series(X1.columns).apply(lambda x : X1[x].dtype == 'O')] 
for i in X1.columns:
    if i in object_col:
        X1[i] = le.fit_transform(X1[i])
    else:
        X1[i] = X1[i]
In [98]:
X_temp, X_test1, y_temp, y_test1 = train_test_split(X1, y1, test_size=0.2, random_state=1, stratify=y1)
In [99]:
X_train1, X_val1, y_train1, y_val1 = train_test_split(X_temp, y_temp, test_size=0.25, random_state=1, stratify=y_temp)
In [100]:
dt.fit(X_train1, y_train1)
y_pred = dt.predict(X_test1)

tn, fp, fn, tp = confusion_matrix(y_test1, y_pred).ravel()

dt_score = dt.score(X_test1, y_test1)
dt_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
dt_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
dt_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['DecisionTree'], 'Accuracy' : [dt_score], 'Precision' : [dt_precision],
                      'True positive rate' : [dt_recall], 'True negative rate' : [dt_specificity],
                       'False positive rate' :  [1-dt_specificity]})

result
Out[100]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 DecisionTree 0.571304 0.874 0.893 0.895 0.105
In [101]:
lr.fit(X_train1, y_train1)

pred = lr.predict(X_test1)
tn, fp, fn, tp = confusion_matrix(y_test1,pred).ravel()

lr_score = lr.score(X_test1, y_test1)
lr_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
lr_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
lr_specificity =round(tn/(tn+fp),3)    # Of all the people without Chds, how many were recognised to not have Chd

result = pd.DataFrame({'Model' : ['Logistic Regression'], 'Accuracy' : [lr_score], 'Precision' : [lr_precision],
                      'True positive rate' : [lr_recall], 'True negative rate' : [lr_specificity],
                       'False positive rate' :  [1-lr_specificity]})
result
Out[101]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 Logistic Regression 0.834711 0.824 0.804 0.86 0.14
In [102]:
rf.fit(X_train1, y_train1)
pred = rf.predict(X_test1)

tn, fp, fn, tp = confusion_matrix(y_test1, pred).ravel()

rf_score = rf.score(X_test1, y_test1)
rf_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
rf_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
rf_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['RandomForest'], 'Accuracy' : [rf_score], 'Precision' : [rf_precision],
                      'True positive rate' : [rf_recall], 'True negative rate' : [rf_specificity],
                       'False positive rate' :  [1-rf_specificity]})
result
Out[102]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 RandomForest 0.92562 0.91 0.926 0.925 0.075
In [103]:
gb.fit(X_train1, y_train1)
pred = gb.predict(X_test1)

tn, fp, fn, tp = confusion_matrix(y_test1, pred).ravel()

gb_score = gb.score(X_test1, y_test1)
gb_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
gb_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
gb_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['GradientBoosting'], 'Accuracy' : [gb_score], 'Precision' : [gb_precision],
                      'True positive rate' : [gb_recall], 'True negative rate' : [gb_specificity],
                       'False positive rate' :  [1-gb_specificity]})
result
Out[103]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 GradientBoosting 0.940771 0.927 0.942 0.94 0.06
In [104]:
xg.fit(X_train1, y_train1)

pred = xg.predict(X_test1)

tn, fp, fn, tp = confusion_matrix(y_test1, pred).ravel()

xg_score = xg.score(X_test1, y_test1)
xg_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
xg_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
xg_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['XGBClassifier'], 'Accuracy' : [xg_score], 'Precision' : [xg_precision],
                      'True positive rate' : [xg_recall], 'True negative rate' : [xg_specificity],
                       'False positive rate' :  [1-xg_specificity]})
result
Out[104]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 XGBClassifier 0.760331 0.68 0.88 0.662 0.338

F. Which data has better performing models, its Original data or Up sampled data or Down Sampled data? Share your insights. [1 Mark]

The recall or True positive rate for Down sampled data
XGBClassifier - 0.896
GradientBoosting - 0.954
RandomForest - 0.939
Logistic Regression - 0.794
DecisionTree - 0.905

The recall or True positive rate for Up sampled data
XGBClassifier - 0.809
GradientBoosting - 0.938
RandomForest - 0.886
Logistic Regression - 0.64
DecisionTree - 0.837

The recall or True positive rate for Original data
XGBClassifier - 0.837
GradientBoosting - 0.89
RandomForest - 0.847
Logistic Regression - 0.552
DecisionTree - 0.813

Interms of recall, Down sampling of data provided best results

G. From the above built models, which 2 models are outperforming with better recall score. Mention and share your insights

Comparing above five models, 1) GradientBoosting and 2) RandomForest algorithms outperform other 3 algorithms for this data.

9. Model Selection and Hyper parameter Tuning [14 Marks]¶

A. Choose any 2 best performing models from above and tune the hyper parameters of those models. [4 Marks]

In [105]:
rf=RandomForestClassifier( bootstrap = False,
  max_depth = 20,
  max_features =  'auto' ,
  min_samples_leaf = 1,
  min_samples_split = 2,
  n_estimators = 200, random_state= 1)

The hyperparameters considered for random forest classifier are max_depth, min_samples_split, n_estimators

In [106]:
gb=GradientBoostingClassifier(n_estimators=500,learning_rate=0.05,random_state=100,max_features=5 )

The hyperparameters considered for gradient boosting classifier are n_estimators, learning_rate,max_features

B. Build the models on train data. [4 Marks]
C. Now validate the models on test data. Check with all performance matrix. Share your insights. [4 Marks]

In [107]:
gb.fit(X_train1, y_train1)
pred = gb.predict(X_test1)

tn, fp, fn, tp = confusion_matrix(y_test1, pred).ravel()

gb_score = gb.score(X_test1, y_test1)
gb_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
gb_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
gb_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['GradientBoosting'], 'Accuracy' : [gb_score], 'Precision' : [gb_precision],
                      'True positive rate' : [gb_recall], 'True negative rate' : [gb_specificity],
                       'False positive rate' :  [1-gb_specificity]})
result
Out[107]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 GradientBoosting 0.940771 0.927 0.942 0.94 0.06
In [108]:
gb=GradientBoostingClassifier(n_estimators=850,learning_rate=0.2,random_state=100,max_features=7 )
In [109]:
 gb.fit(X_train1, y_train1)
 pred = gb.predict(X_test1)

 tn, fp, fn, tp = confusion_matrix(y_test1, pred).ravel()

 gb_score = gb.score(X_test1, y_test1)
 gb_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
 gb_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
 gb_specificity =round(tn/(tn+fp),3) 

 result = pd.DataFrame({'Model' : ['GradientBoosting'], 'Accuracy' : [gb_score], 'Precision' : [gb_precision],
                      'True positive rate' : [gb_recall], 'True negative rate' : [gb_specificity],
                       'False positive rate' :  [1-gb_specificity]})
 result
Out[109]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 GradientBoosting 0.949036 0.934 0.954 0.945 0.055
In [110]:
rf.fit(X_train1, y_train1)
pred = rf.predict(X_test1)

tn, fp, fn, tp = confusion_matrix(y_test1, pred).ravel()

rf_score = rf.score(X_test1, y_test1)
rf_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
rf_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
rf_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['RandomForest'], 'Accuracy' : [rf_score], 'Precision' : [rf_precision],
                      'True positive rate' : [rf_recall], 'True negative rate' : [rf_specificity],
                       'False positive rate' :  [1-rf_specificity]})
result

C:\Users\SUBASHINI\anaconda3\lib\site-packages\sklearn\ensemble\_forest.py:424: FutureWarning: `max_features='auto'` has been deprecated in 1.1 and will be removed in 1.3. To keep the past behaviour, explicitly set `max_features='sqrt'` or remove this parameter as it is also the default value for RandomForestClassifiers and ExtraTreesClassifiers.
  warn(
Out[110]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 RandomForest 0.932507 0.918 0.933 0.932 0.068
In [111]:
rf=RandomForestClassifier( bootstrap = False,
  max_depth = 40,
  max_features =  'auto' ,
  min_samples_leaf = 1,
  min_samples_split = 3,
  n_estimators =250, random_state= 1)
In [112]:
rf.fit(X_train1, y_train1)
pred = rf.predict(X_test1)

tn, fp, fn, tp = confusion_matrix(y_test1, pred).ravel()

rf_score = rf.score(X_test1, y_test1)
rf_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
rf_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
rf_specificity =round(tn/(tn+fp),3) 

result = pd.DataFrame({'Model' : ['RandomForest'], 'Accuracy' : [rf_score], 'Precision' : [rf_precision],
                      'True positive rate' : [rf_recall], 'True negative rate' : [rf_specificity],
                       'False positive rate' :  [1-rf_specificity]})
result

C:\Users\SUBASHINI\anaconda3\lib\site-packages\sklearn\ensemble\_forest.py:424: FutureWarning: `max_features='auto'` has been deprecated in 1.1 and will be removed in 1.3. To keep the past behaviour, explicitly set `max_features='sqrt'` or remove this parameter as it is also the default value for RandomForestClassifiers and ExtraTreesClassifiers.
  warn(
Out[112]:
Model Accuracy Precision True positive rate True negative rate False positive rate
0 RandomForest 0.932507 0.924 0.926 0.938 0.062

10. Conclusion [4 Marks]¶

A. Pickle the selected model for future use. [2 Marks]

In [113]:
def gbfit_tune():
 gb=GradientBoostingClassifier(n_estimators=500,learning_rate=0.05,random_state=100,max_features=5 )
 gb.fit(X_train1, y_train1)
 pred = gb.predict(X_test1)

 tn, fp, fn, tp = confusion_matrix(y_test1, pred).ravel()

 gb_score = gb.score(X_test1, y_test1)
 gb_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
 gb_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
 gb_specificity =round(tn/(tn+fp),3) 

 result = pd.DataFrame({'Model' : ['GradientBoosting'], 'Accuracy' : [gb_score], 'Precision' : [gb_precision],
                      'True positive rate' : [gb_recall], 'True negative rate' : [gb_specificity],
                       'False positive rate' :  [1-gb_specificity]})
 result
In [114]:
def rffit_tune():
 rf.fit(X_train1, y_train1)
 pred = rf.predict(X_test1)

 tn, fp, fn, tp = confusion_matrix(y_test1, pred).ravel()

 rf_score = rf.score(X_test1, y_test1)
 rf_recall = round(tp/(tp+fn), 3)   # Of all the people with Chd, how many were recognised to have Chd
 rf_precision = round(tp/(tp+fp), 3)  # Of all the people predicted to have Chd, how many did have Chd
 rf_specificity =round(tn/(tn+fp),3) 

 result = pd.DataFrame({'Model' : ['RandomForest'], 'Accuracy' : [rf_score], 'Precision' : [rf_precision],
                      'True positive rate' : [rf_recall], 'True negative rate' : [rf_specificity],
                       'False positive rate' :  [1-rf_specificity]})
 result
In [115]:
def main():
    gbfit_tune()
    rffit_tune()
In [116]:
import dill as pickle
g = pickle.dumps(main)
In [117]:
pickle.loads(g)()

C:\Users\SUBASHINI\anaconda3\lib\site-packages\sklearn\ensemble\_forest.py:424: FutureWarning: `max_features='auto'` has been deprecated in 1.1 and will be removed in 1.3. To keep the past behaviour, explicitly set `max_features='sqrt'` or remove this parameter as it is also the default value for RandomForestClassifiers and ExtraTreesClassifiers.
  warn(

B. Articulate some Business Recommendations and observation from the model. [2 Marks]

Having considered various evaluation metrics , we can conclude that the gradient boosting classifier is the best performing algorithm for our data amongst the other explored models to study and predict customer churn behaviour.
This model can help the bank predict which customers might consider leaving. With such information, the bank can proactively prevent the customer from leaving by addressing the issues the customers might be facing that incentivized them to consider leaving.

In [ ]: